Relating the Bures measure to the Cauchy two-matrix model (Q5962908): Difference between revisions

The Bures metric (Helstrom metric) defines an infinitesimal distance between density matrix operators representing the quantum states. Actually, the distance between two states can be represented as a finite version of the infinitesimal square distance. Then the Bures distance between two states can be defined by the equality: \[ D_B(\rho_1 , \rho_2 )^2 = 2 (1 - \sqrt{F ( \rho_1,\rho_2 )} \;) , \] where \(F\) is the fidelity function. Meanwhile, it can be described as a quantum generalization of the Fisher information metric. The authors use the following formula: \[ D_B (\rho_1,\rho_2 )= \sqrt{2-2 \text{tr}\sqrt{\sqrt{\rho_1}\rho_2\sqrt{\rho_1}}}. \] The Cauchy two-matrix model was introduced by \textit{M. Bertola} et al. [Commun. Math. Phys. 287, No. 3, 983--1014 (2009; Zbl 1197.82037)]. This model includes two-matrix ensemble, that is a measure on two \(N\times N\) positive definite Hermitian matrices \(M_1\), \(M_2\) distributed by a very complex formula given in the paper. A relation with this model was only introduced but never thoroughly studied. In the present paper the authors introduce general partition functions for the Cauchy two-matrix ensemble as for this purpose use an integral over products and ratios of characteristic polynomials weighted by the joint probability density. The bi-orthogonal polynomials as well their Cauchy transforms, and other transforms are discussed. These polynomials along with the partition functions comprising of two characteristic polynomials are used to determine the kernels of the determinantal point process. It turns out that the explicit form of the \((k, l)\)-point correlation function of the joint probability density can be analyzed. The proof of these results are briefly sketch. It follows from the paper that any large \(N\) asymptotics of the level statistics including the macroscopic and microscopic level densities can be analyzed. Also the hard edge and the soft edge correlation functions, and the correlation functions in the bulk could be of interest.